Dynamic Bayesian Networks for Evaluation of Granger Causal Relationships in Climate Reanalyses. (1st May 2021)

• Record Type: Journal Article
• Authors: Harries, Dylan; O'Kane, Terence J.
• Is Part Of: Journal of advances in modeling earth systems. Volume 13:Number 5(2021)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-05-01
• Subjects: Bayesian -- graphical models -- teleconnections; Geological modeling -- Periodicals; Climatology -- Periodicals; Geochemical modeling -- Periodicals; 551.5011
• DOI: 10.1029/2020MS002442
• Languages: English
• ISSNs: 1942-2466

Abstract

Abstract: We apply a Bayesian structure learning approach to study interactions between global climate modes, so illustrating its use as a framework for developing process‐based diagnostics with which to evaluate climate models. Homogeneous dynamic Bayesian network models are constructed for time series of empirical indices diagnosing the activity of major tropical, Northern and Southern Hemisphere modes of climate variability in the NCEP/NCAR and JRA‐55 reanalyses. The resulting probabilistic graphical models are comparable to Granger causal analyses that have recently been advocated. Reversible jump Markov Chain Monte Carlo is employed to provide a quantification of the uncertainty associated with the selection of a single network structure. In general, the models fitted from the NCEP/NCAR reanalysis and the JRA‐55 reanalysis are found to exhibit broad agreement in terms of associations for which there is high posterior confidence. Differences between the two reanalyses are found that involve modes for which known biases are present or that may be attributed to seasonal effects, as well as for features that, while present in point estimates, have low overall posterior mass. We argue that the ability to incorporate such measures of confidence in structural features is a significant advantage provided by the Bayesian approach, as point estimates alone may understate the relevant uncertainties and yield less informative measures of differences between products when network‐based approaches are used for model evaluation. Plain Language Summary: To produce reliable forecasts and projections, climate models should accurately reproduce the observed behavior of different processes that play a role in Earth's climate, including the relationships between them. Statistical methods can be used to describe these interactions in models and in observations, and which can then be compared to evaluate how well a given model captures the observed relationships. However, networks obtained from estimates of the true historical state of the climate, known as reanalyses, will also be affected by the properties of the systems used to create these estimates, as well as random variability, and hence may have significant uncertainties. Using what are known as Bayesian statistical methods, we estimate the uncertainties associated with particular interactions in two widely used reanalyses. Interactions that are found to be very likely to be present in one reanalysis but not the other are suggested to be due to systematic differences in the two reanalysis systems and need to be kept in mind when these state estimates are used to evaluate climate models. Therefore, it is important to account for the uncertainty associated with each relationship when analyzing state estimates and further employing them to evaluate climate models. Key Points: Bayesian structure learning provides a principled approach to quantifying uncertainty in estimated network structures for relationships between climate modes Dynamic Bayesian networks estimated from NCEP/NCAR and JRA‐55 reanalysis data show broad overall consistency Structural differences in high posterior credibility associations may be indicative of biases relevant for subsequent model evaluation